AI, Indigenous Language Sovereignty, and Decolonizing Technology
Full Show Notes: https://www.theintelligenceagepodcast.com/846
In this conversation, Mark Smth speaks with Natasha MacDonald, an Inuit educator and Heritage Lab volunteer, about protecting Indigenous language and culture in the age of AI. Natasha explains how her community is using closed-loop technology, community vetting, and local control to preserve Inuktitut without handing data to big tech.
We also discuss colonial language loss, intellectual property, youth-led revitalization, and why cultural data should stay under Indigenous sovereignty.
Key topics
Natasha shares her Inuit background, her family roots in Nunavik, and how food, travel, and cross-cultural experience shaped her perspective.
She explains the history of colonization in her region, including missionaries, federal day schools, forced settlement, and the shift from nomadic life to permanent communities.
Natasha describes how Inuktitut remains strong in Nunavik, with roughly 95% of the local Inuit population still speaking or using it, even as English pressure grows.
We discuss Heritage Lab, where Natasha serves as vice president, and its mission to protect and preserve Indigenous languages through technology.
She explains the nonprofit’s closed-loop model, including community-owned data, language committee approval, and hosting on servers located on Indigenous lands.
Natasha outlines the risks of big tech and external institutions scraping language data, training AI models without consent, and turning Indigenous knowledge into profit.
She describes a dispute involving a University of Laval partnership with Anthropic and why her community objected to Inuit data being used to train AI on low-resource languages.
We discuss the difference between data sovereignty and intellectual property, and how Indigenous knowledge is treated as community-held rather than individually owned.
Natasha emphasizes that language revitalization must include youth, who are building tools, games, and models while learning pride and agency in the process.
She also talks about the role of art as cultural memory, economic support, and a point of resistance against AI-generated imitation.
Resources:
Heritage Lab: https://heritagelab.ca/en/story
Mila: https://mila.quebec/en/news/mila-and-cifar-launch-the-indigenous-ai-advisory-circle
If you want to get in touch with me, you can message me here on Linkedin.
Thanks for listening 🚀 – Mark Smith
00:00 Welcome and introduction. Mark introduces Natasha and how the conversation came about.
00:57 Food, family and fun. Natasha on Inuit culture, her mixed heritage and growing up between cultures.
02:40 Language loss and AI. Mark shares his connection through his Māori wife and the questions AI raises.
03:15 Inuit of Nunavik. Natasha describes her community, its place in Canada and her territory.
04:50 Colonization across generations. Her grandmother's life on the land, then federal day schools, and the pressure of English and French.
08:06 Language today. About 95% still speak Inuktitut, but youth are shifting to English.
09:34 Heritage Lab. Natasha's work protecting Indigenous languages with closed-loop AI and community-vetted data.
11:50 Keeping data sovereign. Mark asks how to stop the data ending up with big tech. Natasha explains decentralized servers on Indigenous land and their small footprint.
14:46 Big tech and the Laval University grant. Attempts to scrape Inuit language, and why the community says no.
19:15 Research, IP and who owns knowledge. Profiting from documented culture, and the Māori concept of community-held knowledge.
21:07 Resource extraction. Furs then, language now. Natasha covers the Calls to Action, UNDRIP and the fake accounts in community groups.
25:08 Urgency. Colonization is recent, and a language can be lost in one generation.
26:57 The next generation and identity. Mark on Te Reo Māori revitalization. Natasha on avoiding homogenization.
30:59 Dialects, and youth building the tools. Reflecting every dialect and hiring Indigenous youth to build the games and models.
34:04 Funding and infrastructure. The race for small grants, and the "McDonald's franchise" model for communities.
36:40 Supporting communities with few speakers. A British Columbia community with only about ten speakers left.
37:46 Art, culture and AI. Family artwork, artists' livelihoods, and the youth backlash against AI-generated imagery.
41:21 Closing thoughts. Parallels with New Zealand, the trade-off of short-term cost against long-term identity, and thank-yous.